aws.dd_forwarder.logs_failed and metrics_failed are never emitted due to passing list instead of len() to send_event_metric

[patbos]

Bug

In aws/logs_monitoring/forwarder.py, send_event_metric is called with the failed items list as the metric value instead of its length:

# Line 141 - passes list, should be len()
send_event_metric("logs_failed", failed_logs)

# Line 172 - passes list, should be len()
send_event_metric("metrics_failed", failed_metrics)

Compare with the correct pattern used on the adjacent lines:

send_event_metric("logs_forwarded", len(logs_to_forward) - len(failed_logs))
send_event_metric("metrics_forwarded", len(metrics) - len(failed_metrics))

Root cause

send_event_metric calls lambda_metric, which validates the value with float(value). Passing a list raises TypeError, which is caught and logged as a warning, then the metric is silently dropped:

[WARNING] Ignoring metric submission for metric 'aws.dd_forwarder.logs_failed'
because the value is not numeric: [...]

See datadog_lambda/metric.py:

try:
    float(value)
except (ValueError, TypeError):
    logger.warning(
        "Ignoring metric submission for metric '%s' because the value is not numeric: %r",
        metric_name,
        value,
    )
    return

Impact

aws.dd_forwarder.logs_failed and aws.dd_forwarder.metrics_failed are never emitted, even when failures are actively occurring. Any monitors or dashboards based on these metrics will never fire.

Fix

send_event_metric("logs_failed", len(failed_logs))
send_event_metric("metrics_failed", len(failed_metrics))

Version

Reproduced on v5.4.3 (and present in prior versions based on git history).